Adaptive Wide Angle Image Distortion Correction and Panoramic Stitching System

• Main Authors: Yu-Sheng Liang, 梁右昇
• Other Authors: 吳崇賓
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/cgi-bin/gs32/gsweb.cgi/login?o=dnclcdr&s=id=%22107NCHU5441039%22.&searchmode=basic

碩士 === 國立中興大學 === 電機工程學系所 === 107 === This thesis, adaptive wide angle image distortion calibration and panoramic stitching system is proposed. The provided system can calibrate image distortion, construct panoramic video and remove the ghost and the seam effect. The proposed methods are divided into three steps: wide angle image distortion calibration、match point searching and video blending. For the steps of wide-angle image distortion calibration, this thesis propose chessboard based calibration which can make distortion image correction have adaptive calibration. For the search of matching points, this thesis proposes to improve the quality of the image on ambient luminance through an adaptive threshold value. The proposed maximum adjacent edge quantity, set the point with the strongest feature information as the reference point to reduce the processing time. In the image fusion, this thesis proposes an adaptive image fusion method, using the density of the edge of the block, Linear weight fusion and selective fusion are used for smooth areas and non-smooth areas respectively. Experimental results, show the adaptive correction for distortion image can suitable for correction at different wide angle. For the surround image, the ghost and the seam are effectively eliminated in the image overlapping area, and the image quality PSNR is 3.67% higher than the full-use linear weight fusion method.